---
license: other
task_categories:
- image-to-text
- optical-character-recognition
tags:
- funsd
- docld
- form-understanding
- ocr
---

# FUNSD benchmark (DocLD)

Test set for the [FUNSD](https://guillaumejaume.github.io/FUNSD/) benchmark used with [DocLD](https://docld.com): 50 form images and 50 ground-truth annotation JSON files.

- **images/** — 50 PNG form images (FUNSD test set).
- **annotations/** — 50 JSON files (FUNSD format: entities with text, box, label).

Code and full benchmark: [github.com/Doc-LD/funsd-bench](https://github.com/Doc-LD/funsd-bench).